Please don't confuse the 'vector' in 'vector processors' (such as some Crays) with the 'vector' in things like 'AltiVec'.
Real vector processors act on dozens of elements simultaneously, e.g. Crays had 64 64-bit elements in their vectors, some architectures had up to 256 elements that would be acted on in parallel. A measly 4 32-bit elements, or similar, is pathetic, and an insult to real vector processing.
"In 2004 a graphics card used a fixed function unit"
References, or retract. General purpose processing units have been used in graphics acceleration since the 80s, and constantly since.
References? Look up every family of DSP or embedded core in Wikipedia, and look at the uses for it.
I bet you're the kind of fool who thinks that there aren't 6 ARM cores on my mobile phone (where the sales material only lists one, the main application processor).
TI never did "vector processors" as such. They did harvard architecture DSPs which would permit streams of data to be squirted in and results to squirted out with very high throughput, even if the latency wasn't brilliant.
They, however, clearly did 'GPU's. The 34010 was a perfect example of that. However, back in the day we still just called them 'DSP's, as we knew that the digital signals being processed were, most of the time, graphics. Except when they were audio, but you tended to use different DSPs for that (namely the 56k family rather than the 34k family).
Well, there's a flaw in your premises. The GPUs aren't designed to accelerate 3d rendering, transformation, lighting, and rasterisations.
They're deisgned to do large quantities of low-precision arithmetic operations on streams of numeric input.
Being able to do *that* enables them to accelerate 3d rendering. But it also enables them to accelerate video compression. And audio processing. And facial recognition. And asymmetric cryptographic primitives. And...
Your humble opinion is clearly correct. It was obvious the first time video cards gained any kind of computational power. How about 1985 with TI's 34010 chip?
However, your frustration in 1996 was misplaced. Back in 1994/1995 I was *doing* this (or equivalent - H.261 video conferencing) with TI's C80 MVP processor (1 RISC master processor, 4 almost-VLIW DSP cores, and a SDMA unit, all on the same die.)
Similar things have been done in the UK in the past. Under some freedom of information law, if you are caught on a CCTV camera, the authority responsible has to provide you with a copy of what was recorded. Comedian Mark Thomas instigated a plan to basically spam CCTVs and the authorities. The DMCA's a much more worthy target than CCTVs, IMHO.
You want no caching at all? You want 'save page source' to not be possible? You want it impossible to download anything off the web? You want it to be impossible for external document viewers to view temporary files? You want there to be no persistent cookie storage? You want there to be no persistent storage of any settings that could be changed from within the program? You want there to be no possibility for plugins?
Your hypothesised browser would apparently have a user-base of at most 1 with that feature-set.
"Why would you want a third core which runs at a fraction the speed of the other two"
Presumably as it's a simpler core with fewer optional modules (jazelle, (vector) floating point, thumb modes,...), with a much lower transister count, smaller cache, less leakage, and lover voltage OPPs, so that at 600MHz it comsumes less than 600/1500 of what the big cortex cores use.
Although in reality most smartphones will have about 5 ARM cores in them. There really aren't that many ASICs as there used to be, many are just general purpose processors, perhaps with that tell-tale sign - 'firmware'. Your wireless card - I bet you there's an ARM core in it. Your bluetooth chip too? Ditto.
Regarding your 'I am not aware...', here's a powertop summary of an mp3 playing on high end ARM-powered mobile phone with no other apps active:
Freq: 500MHz 3.4%, 250MHz 96.6% IRQs in 30s: DMA 3200, [elided] 2150, programable timer 1880, i2c 640 Core power domain: Off 0%, Retention 0%, Inactive 48%, On 51% (other powrdomains like graphics-related stuff: 100% off) Total wake-ups 10000 = 333/s, Total IRQs 8000 = 266/s, Timers 2000 = 66/s H/W wakeups 100 = 3.3/s
(OK that was on prototype hardware, but not significantly different from the production version. The kernel was an R&D kernel too, so might have additional logging overhead, but I don't think so, however, I wasn't running it on a jig, it was a live system (i.e. my phone).)
So I guess you do now know of one.
Just for reference, this device will play mp3s for longer than a fully-charged iPhone (or what Apple claim in their marketting bumf). So I'm guessing Apple have an even less optimal architecture.
Note to $EMPLOYER - this is on the device that's already on the market, anyone could have obtained these results.
Not just that, but they have two methods of achieving that - if they can't build something to beat the competition, they'll _buy_ something to beat the competition.
[hanging head in shame about similarities to own $DAYJOB - don't blame me, I'm just one of the grunts]
Day _admitted_ that he made a mistake, both b reversng it, and later apologetically in writing. No matter how nice and honest he is, perhaps because of that, he's clearly manipulable. And Mitchell is clearly assertive, so could exploit that with things less high tech than real Jedi mind tricks. Day probably trusted Mruczek (or whatever his name was, I'm not familiar with the characters, and simply downloaded the movie for a quick view this evening) too much too. I'd like to hear more about the resignation.
There needs to be an 11th option alongside the scores - 'turned it off'. Neutral when it comes to numerics, honest when it comes to statistics, and still carrying some information.
The _telco_ world was never one of the 1024 users. Telecomms is all based around the old bitrates of the telephone systems, which were always multiples of 1000 bits per second:
Historically, audio telephony had a sampling frequency 8 kHz.
As we went digital, G711 audio channels and ISDN B channels were 64000 b/s for 8-bit audio. Other codecs shrank that to 32000 b/s, 16000 b/s, etc. ISDN D channels were 16000 b/s. ISDN PRI channels were variously 1544000 b/s or 2048000 b/s. All the fat pipes carrying data around, be that SONET (PDH) or SDH use variously 8448000 b/s, 34368000 b/s, 139264000 b/s, etc., etc.
All those numbers are multiples of the original 8 kHz.
So don't blame telephony for the 1024s - they're the least guilty. (Telephony would even include multipliers like 30 or 31, as you'd do the old power-of-2 thing, then reserve space for control or stuffing, or...)
I contend that we are both atheists. I just believe in one fewer god than you do. When you understand why you dismiss all the other possible gods, you will understand why I dismiss yours. -- Stephen Roberts
Slashdot Mirror
The TI C80MVP existed back in the early 90s, and had everything that's on a modern GPU, just not in the same quantities.
Please don't confuse the 'vector' in 'vector processors' (such as some Crays) with the 'vector' in things like 'AltiVec'.
Real vector processors act on dozens of elements simultaneously, e.g. Crays had 64 64-bit elements in their vectors, some architectures had up to 256 elements that would be acted on in parallel. A measly 4 32-bit elements, or similar, is pathetic, and an insult to real vector processing.
"In 2004 a graphics card used a fixed function unit"
References, or retract. General purpose processing units have been used in graphics acceleration since the 80s, and constantly since.
References? Look up every family of DSP or embedded core in Wikipedia, and look at the uses for it.
I bet you're the kind of fool who thinks that there aren't 6 ARM cores on my mobile phone (where the sales material only lists one, the main application processor).
TI never did "vector processors" as such. They did harvard architecture DSPs which would permit streams of data to be squirted in and results to squirted out with very high throughput, even if the latency wasn't brilliant.
They, however, clearly did 'GPU's. The 34010 was a perfect example of that. However, back in the day we still just called them 'DSP's, as we knew that the digital signals being processed were, most of the time, graphics. Except when they were audio, but you tended to use different DSPs for that (namely the 56k family rather than the 34k family).
Well, there's a flaw in your premises. The GPUs aren't designed to accelerate 3d rendering, transformation, lighting, and rasterisations.
...
They're deisgned to do large quantities of low-precision arithmetic operations on streams of numeric input.
Being able to do *that* enables them to accelerate 3d rendering. But it also enables them to accelerate video compression. And audio processing. And facial recognition. And asymmetric cryptographic primitives. And
Your humble opinion is clearly correct. It was obvious the first time video cards gained any kind of computational power. How about 1985 with TI's 34010 chip?
However, your frustration in 1996 was misplaced. Back in 1994/1995 I was *doing* this (or equivalent - H.261 video conferencing) with TI's C80 MVP processor (1 RISC master processor, 4 almost-VLIW DSP cores, and a SDMA unit, all on the same die.)
Similar things have been done in the UK in the past. Under some freedom of information law, if you are caught on a CCTV camera, the authority responsible has to provide you with a copy of what was recorded. Comedian Mark Thomas instigated a plan to basically spam CCTVs and the authorities. The DMCA's a much more worthy target than CCTVs, IMHO.
"Mmm, about as valuable as my collection of 5 1/4 inch floppies."
:-D
Imagine how valuable *these* will be: http://fatphil.org/how_floppy.jpg
I'm going to be rich, I tell you!
But Darwinian evolutions's nothing to do with staying alive, it's about propagating your genes.
Yes it's semantic - all communication is. Without paying attention to semantics, there's no point in communicating.
I'm clearly too dense to understand that when you say "X" you mean "not X but something different".
And I'd rather stay that way, thanks.
"no browser process can access my disk"
You want no caching at all?
You want 'save page source' to not be possible?
You want it impossible to download anything off the web?
You want it to be impossible for external document viewers to view temporary files?
You want there to be no persistent cookie storage?
You want there to be no persistent storage of any settings that could be changed from within the program?
You want there to be no possibility for plugins?
Your hypothesised browser would apparently have a user-base of at most 1 with that feature-set.
Are you too freaking stupid to understand what the Darwin awards are? (Rhetorical question - the answer's clearly "yes".)
He's bred - he's already passed his genes on to future generations.
Double coincidence - "Jimi was chairman of Hesco Bastion Ltd, the world leading manufacturer of protective barriers"
Hmmm, he didn't sell enough, clearly.
"Why would you want a third core which runs at a fraction the speed of the other two"
...), with a much lower transister count, smaller cache, less leakage, and lover voltage OPPs, so that at 600MHz it comsumes less than 600/1500 of what the big cortex cores use.
Presumably as it's a simpler core with fewer optional modules (jazelle, (vector) floating point, thumb modes,
But that's a WSITD.
You're behind the curve, the cell had 9 cores.
Although in reality most smartphones will have about 5 ARM cores in them. There really aren't that many ASICs as there used to be, many are just general purpose processors, perhaps with that tell-tale sign - 'firmware'. Your wireless card - I bet you there's an ARM core in it. Your bluetooth chip too? Ditto.
Regarding your 'I am not aware ...', here's a powertop summary of an mp3 playing on high end ARM-powered mobile phone with no other apps active:
Freq: 500MHz 3.4%, 250MHz 96.6%
IRQs in 30s: DMA 3200, [elided] 2150, programable timer 1880, i2c 640
Core power domain: Off 0%, Retention 0%, Inactive 48%, On 51%
(other powrdomains like graphics-related stuff: 100% off)
Total wake-ups 10000 = 333/s, Total IRQs 8000 = 266/s, Timers 2000 = 66/s
H/W wakeups 100 = 3.3/s
(OK that was on prototype hardware, but not significantly different from the production version. The kernel was an R&D kernel too, so might have additional logging overhead, but I don't think so, however, I wasn't running it on a jig, it was a live system (i.e. my phone).)
So I guess you do now know of one.
Just for reference, this device will play mp3s for longer than a fully-charged iPhone (or what Apple claim in their marketting bumf). So I'm guessing Apple have an even less optimal architecture.
Note to $EMPLOYER - this is on the device that's already on the market, anyone could have obtained these results.
Not just that, but they have two methods of achieving that - if they can't build something to beat the competition, they'll _buy_ something to beat the competition.
[hanging head in shame about similarities to own $DAYJOB - don't blame me, I'm just one of the grunts]
You score less and less as time progresses, but you do get extra lives?
Did you miss the "Jedi mind tricks" line?
Day _admitted_ that he made a mistake, both b reversng it, and later apologetically in writing. No matter how nice and honest he is, perhaps because of that, he's clearly manipulable. And Mitchell is clearly assertive, so could exploit that with things less high tech than real Jedi mind tricks. Day probably trusted Mruczek (or whatever his name was, I'm not familiar with the characters, and simply downloaded the movie for a quick view this evening) too much too. I'd like to hear more about the resignation.
There needs to be an 11th option alongside the scores - 'turned it off'. Neutral when it comes to numerics, honest when it comes to statistics, and still carrying some information.
Doh - that's what you're saying. Tedious facts intended for your parent poster.
The _telco_ world was never one of the 1024 users. Telecomms is all based around the old bitrates of the telephone systems, which were always multiples of 1000 bits per second:
Historically, audio telephony had a sampling frequency 8 kHz.
As we went digital, G711 audio channels and ISDN B channels were 64000 b/s for 8-bit audio. Other codecs shrank that to 32000 b/s, 16000 b/s, etc. ISDN D channels were 16000 b/s. ISDN PRI channels were variously 1544000 b/s or 2048000 b/s. All the fat pipes carrying data around, be that SONET (PDH) or SDH use variously
8448000 b/s, 34368000 b/s, 139264000 b/s, etc., etc.
All those numbers are multiples of the original 8 kHz.
So don't blame telephony for the 1024s - they're the least guilty. (Telephony would even include multipliers like 30 or 31, as you'd do the old power-of-2 thing, then reserve space for control or stuffing, or...)
Your gas stove isn't, and can not be, gas either.
And I'm pretty sure the last french fry you ate wasn't French.
Please attempt to grasp a few more of the many ways in which compound nouns may be formed in English.
"The best way to bring the birthrate down, though, is to improve the economic and legal status of women throughout the world."
I'd have thrown education in there too. (And hopefully with education, parity of economic status would be more closely approached.)
From my .sig file, this seems appropriate:
I contend that we are both atheists. I just believe in one fewer god
than you do. When you understand why you dismiss all the other possible
gods, you will understand why I dismiss yours. -- Stephen Roberts